Understanding PyTorch

00:00 In this lesson, you’ll get a general idea of PyTorch and delve into its code style and ecosystem. Let’s start by exploring some of its general features. The name PyTorch is derived from PI for Python and Torch, which is the backend of PyTorch.

00:17 You’ll be learning a lot more about Torch in the next few slides.

00:22 PyTorch was developed by Facebook AI Research Lab, or FAIR, in 2016. FAIR is now known as the Fundamental AI Research by Meta.

00:36 PyTorch is more popular in research because of its Python nature, ease of debugging, and tight integration with other AI tools like Anaconda, which is a popular Python distribution for scientific computing.

00:51 And NVIDIA CUDA, which is a parallel computing platform and application programming interface model.

00:59 This tight integration allows for efficient GPU acceleration, which is crucial for handling large data sets and complex computational tasks commonly encountered in research. PyTorch is almost as popular as TensorFlow as of 2023.

01:17 This graph shows the 2023 Stack Overflow developer survey list of most popular other frameworks, libraries, and tools, which reports that 8.41% of developers use TensorFlow and 7.89% use PyTorch.

01:36 That’s pretty close for some context of how far PyTorch has come. The 2020 Stack Overflow developer survey reported that 10.4% of professional developers choose TensorFlow and only 4.1% choose PyTorch.

01:54 And in 2018, the percentages were 7.6% for TensorFlow and just 1.6% for PyTorch. PyTorch is definitely catching up.